1. Lecture 3: Contextual Analysis
Brad Myers
/ : Introduction to Human Computer Interaction for Technology Executives
Fall, 2017, Mini 2
© Brad Myers

2. Why Context? Design complete work process Integration!
Fits into “fabric” of entire operations
Not just “point solutions” to specific problems
Integration!
Consistency, effectiveness, efficiency, coherent
Design from data
Not just opinions, negotiation
Not just a list of features
© Brad Myers

3. Key distinctions about CIs
Interviews, Surveys, Focus Groups
Summary data & abstractions
What customers say
Subjective
Limited by reliability of human memory
What customers think they want
Contextual Inquiry
Ongoing experience & concrete data
What users do
Objective
Spontaneous, as it happens
What users actually need
© Brad Myers

4. Who? Users Interviewers: “Cross-functional” team
Between 6 – 20
Representative of different roles
Note: may not be people who will be doing the purchasing of the system
E.g., if for an enterprise; public kiosk
For HW#1 should be novice
Interviewers: “Cross-functional” team
Designers
UI specialists
Product managers
Marketing
Technical people
Generally, at least 2 interviewers
© Brad Myers

5. Some Alternative Contextual Inquiry Interview Methods
For tasks that are:
Intermittent
Uninterruptible
Extremely long
Multi-person tasks
Alternatives to “regular” Contextual Inquiry:
In-context cued recall
Activity logs
Post-observation inquiry
Artifact walkthrough
Retrospective interview
© Brad Myers

6. Contextual Inquiry Issues to observe
Interview of work in progress, in “context”
Actual session of doing a task
Not an interview asking about possible tasks, etc.
Note that focusing on expert behavior & breakdowns
Questions to clarify about routine, motivations
Why do certain actions: need intent for actions
Notice problems (“breakdowns”)
Notice what happens that causes users to do something (“triggers”)
E.g. appearance of error messages, other feedback, external events (phone ringing), etc.
© Brad Myers

7. Example of CI Brand new video of sample session with an eCommerce site
Re-creation of actual activities my wife had with a current website
© Brad Myers

8. Parts of Homework 1 Part A: Part B:
Describe one or more representative tasks
Script of exactly what you will say
Part B:
Description of the user
Transcript of what the user did and said, and what you did and said
Part C: Contextual Analysis diagrams
Artifact model
Flow model
Social model 
Optional: physical model
© Brad Myers

9. Script example from a previous year – Initial Questions
Garmin Nuvi 200 portable automobile GPS by Puja Subramanyam
© Brad Myers

10. Example, cont.: Tasks & User
© Brad Myers

11. Transcript Need to create a transcript of what is said
Can be exact or a summary
Must have time codes or line numbers
© Brad Myers

12. Start of Transcript
No need to follow your script exactly if it doesn’t make sense.
© Brad Myers

13. What to do with all the data?
How organize and use data from contextual inquiry?
Recommended technique: Use Graphical Models (diagrams)
Integrate
Summarize
Point back to the data
Help inspire and guide design
Provide shared focus
Provide an intermediate deliverable outside design team
“Opinions are cheap but insights are priceless.”
© Brad Myers

14. Why Graphical? Displays whole picture at once Reveals patterns
Forces summarization and conciseness
Multiple models provides different perspectives
© Brad Myers

15. Analysis Beyer & Holtzblatt call this process “Contextual Design”
But Hartson & Pyla call it “Analysis”
Better name!
Hartson & Pyla recommend doing a Work Activity Affinity Diagram (WAAD) first, but we go right from data  models
Affinity diagrams are more important for teams and when there are lots of ideas
You will make the models directly from your transcripts
© Brad Myers

16. Models: Flow Model Social / Cultural Model Artifact Model
Communication and coordination necessary to perform task
Social / Cultural Model
Constraints on work due to policy, culture, or values
Artifact Model
Physical things used and created
Physical Model
Layout of work environment as it affects the work.
© Brad Myers

17. More Models
Other models are in Hartson & Pyla but not covered further here (not in homeworks or on exam)
Task decompositions
Usage Scenarios (narratives about tasks)
Essential Use Cases (user and system steps)
Information Object Models (information needed and generated)
Sequence / Step-by-Step Model
Detailed work steps
Sequence in Beyer & Holtzblatt expanded in multiple models
“Step by Step Task Interaction Model”
Providing both “current” and “envisioned” versions of the models
We are only doing the current version
© Brad Myers

18. In ALL Models
ALWAYS label everything you put on the model with the line-number in the transcript (or time on the video/audio, if no transcript).
This allows you to get back to the raw data to see if a different interpretation is possible.
If there is no pointer back, then it is an unsupported Assumption and subject to additional scrutiny.
Don’t make things up – show only what is available in the data
Any assumptions must be marked with (A)
© Brad Myers

19. “Breakdowns” / “Barriers”
Problems that interfere with normal operation
Represented in most models
Key opportunities and issues for design
Often shown in red with a “lightning bolt”
Can’t see the screen
© Brad Myers

20. Additional Example New example: Giving a presentation From our paper:
Karen Cross, Adrienne Warmack, and Brad Myers. "Lessons Learned: Using Contextual Inquiry Analysis to Improve PDA Control of Presentations“
Results influenced design of SlideShow Commander remote control product for mobile phones
Used in the Hartson-Pyla textbook
© Brad Myers

21. Flow Model Communication and coordination necessary to perform tasks
Work flow
Who talks to whom? Who gives what to whom?
Key roles of individuals or groups
Key responsibilities of that person with respect to the tasks
Flow of communication and artifacts, shown as arrows
Artifacts passed around
Actions along the way
Places that things or people go
Breakdowns
© Brad Myers

22. Flow Model structure Note: you (interviewer) are never in any model
Circles and/or icons = people or groups by role
Boxes = things (artifacts), places, files, etc.
Arrows = flow
Red lightening bolts = breakdowns
Times refer to time codes in video
Could also refer to lines of a transcript
For homework, approximate times or line numbers are OK
“(A)” = Assumption = interpretation
Note: not for team-member's opinions about the UI
Instead (A) is for assumptions about what user did
Note: you (interviewer) are never in any model
© Brad Myers

23. Flow Model components General Web
How do job responsibilities get assigned to people?
How do they get help?
How do new tasks get assigned, and how are they carried out?
Coordination: where did each artifact come from and where does it go?
Problems with coordination: forgetting, timing, steps
Creates the “bird’s eye view” of organizational structure
Web
NOT flow chart of pages visited
How information and command flow among the site(s)
© Brad Myers

24. Flow Model Example
© Brad Myers

25. Flow Model Example (SSC)
Fig 6-8 in text
© Brad Myers

26. Social Model Beyer & Holtzblatt call this “Cultural Model”
Culture of organization, family, community defines expectations, desires, policies, values and approach
“Culture is as invisible as water to a fish”
Pervasive, inescapable; yet invisible and intangible
Types of influences:
Formal and informal policies
Power of individuals and groups over each other
Values of company or team
Work domain constraints
Group’s sense of identity
People’s emotions about what they do
The style, values and preferences of individuals or teams
More examples of what goes on social model:
When acceptable to use a recording device
What friends might think
© Brad Myers

27. Social Model Structure
Ovals for “Influencers”: individuals or groups, internal or external
Thought bubbles for feelings/concerns that they actually expressed
Note: often from the preliminary questions or opinions during the CI
Arrows for direction of influence
Labels for samples of dialog showing type of influence and attitudes
Worded as commands
Also show “pushback” – influence in other direction
Breakdowns
In relationships among people
No need to repeat previously shown problems
NOTE: Not allowed to make stuff up!
Just what you actually have data to support!
© Brad Myers

28. Social Model Contents What to put into social model for appliances:
Context of use: when used, other people around, whether OK to record other people, what it looks like
Feelings: proud to own it, embarrassed
Influences: why buy one vs. another, qualities desired
But need data to support all claims
From initial interview questions or other evidence
Can’t just make stuff up!
© Brad Myers

29. Social Model Example

30. Social Model Example (SSC)
Fig 6-6 from text
Typo in book! This arrow goes the wrong way!
© Brad Myers

31. Artifact Model Model = drawing, photograph or copy of real artifact
Artifacts: What people create, modify and use as part of tasks
Reveal traces of people’s work practices
Examples:
Handwritten notes and signatures on hardcopy project plans -> information flow and approval
Fancy formatting on spreadsheet -> looks are important
Screen shots showing problematic features
Model = drawing, photograph or copy of real artifact
Annotate with observations
© Brad Myers

32. Artifact Model Examples
© Brad Myers

33. Artifact Models, cont.
© Brad Myers

34. Physical Model Way the physical environment affects tasks
E.g, placement of items on a desk
Proximity of printers
Can’t hold a device with a keyboard while standing up
In presentation example, where people are and layout of environment
Note: Physical model not always relevant or needed
Seems less important for web, unless mobile
Not required for homework 1, but please do one if environment affects your system
© Brad Myers

35. Components of Physical Model
Places in which work occurs
Physical structures which limit or define the space
Usage and movement within the space
Hardware and other artifacts used
Layout of tools and artifacts
Positions of people within environment
Breakdowns due to physical environment
© Brad Myers

36. Physical Model, example
© Brad Myers

37. Beyer&Holtzblatt’s Sequence Model
Similar to Hartson&Pyla’s “Step-by-Step Task Interaction Model”
Steps taken to complete a task
Triggers that cause the step to happen
E.g., at a particular time; when something else happens
Intent is key to understanding the steps
Also called the goal
Why each step is performed, and why in that order
Arrows to show order of the steps
Can have loops
Breakdowns in communication or coordination
(Note: this model not in homeworks or exam)
© Brad Myers

38. Sequence Model Components
Can choose level of detail depending on focus (what investigating)
E.g., for writing a letter:
High-level (functional level): Find most recent letter written to same person, open it, delete date, replace with new date, delete contents, type new contents, …
Low level (user interface level): Switch windows explorer to details view, Sort files by date, double-click on top item, check “To:” to see if correct person, click and drag across date field, …
Notice hesitations and errors
Interrupt and ask why or what expected
© Brad Myers

39. Hartson&Pyla’s Task Structure Models
Their replacement for Sequence Model
Tasks that need to be supported by the system
You need to decide the important tasks in order to:
Optimize interface & design – what is important?
Design user study tasks – what will participants do?
Understanding tasks can help with better designs because organize UI by task, not by function
What to do, not how
Information needs – what need to know to do task
Unlike sequence models, task models try to capture all requirements, not just the ones in the data
© Brad Myers

40. Hartson&Pyla’s Hierarchical Task Inventory
Hierarchical Task Inventory (HTI) shows tasks and subtasks
Doing a subtask is part of doing the parent task
© Brad Myers

41. Creating Models Create models generalizing over all interviewees
“Consolidated” models
Barrier summaries
Key Idea: Induce generalizations from concrete data
Don’t rely on intuition alone
Don’t deduce from logical abstractions
Example:
Logic says system manager will diagnose the reason behind a system failure. Actual practice: System manager tries standard fixes first (like reboot) & diagnoses only if necessary
Main goal: Deduce the intent
© Brad Myers

42. What To Do With Models User data drives innovation
Solve problems (breakdowns) identified in models
Grounded brainstorming
Flow model
Eliminate flows, roles, redundant data entry
Social / Cultural model
Increase communication, reinforce positive values
Artifact model:
Guide requirements, metaphors, remove screen problems
Physical model:
Depend only on what is available, reduce motion, improve flow of artifacts
Sequence model:
Eliminate, automate steps
© Brad Myers